WeKnora/website-docs/03-features
lyingbug 9aa63c866f
fix(models): correct four vendor request shapes that contradict the docs (#3489)
PR #3470 逐条核对了 27 家厂商,但有四处(五个 entry)的声明与厂商文档对不上,
其中一处在默认路径上就是错的。

1. DeepSeek:参考文档写明 "required and named tool choices are not supported
   in thinking mode; the API returns a 400 error",而 vendor.go 的事实注释写成
   「plain OpenAI set」并因此没有配 tool_choice_modes,默认继承全集。现在只发
   none / auto,和 Moonshot 同类限制的写法一致。compat 无法表达「仅思考模式下
   受限」,而思考是 DeepSeek 的默认状态,所以不做条件区分。

2. 智谱 glm-5.2:官方对 GLM-5.2 的折叠规则是 none|minimal 放弃思考、low|medium
   折成 high。models.json 却把 minimal 改写成 "low",发出去被平台折成 high,
   用户选最弱档实际跑最高强度。删掉该覆盖,回落到厂商级的 minimal。

3-4. NVIDIA NIM:z-ai/glm-5.3 与 moonshotai/kimi-k3 配了 thinking_levels,却
   沿用厂商默认的 chat-template-kwargs 编码——该分支只发 enable_thinking 布尔,
   档位被整个丢弃,而前端仍按 capabilities 渲染出一个不起作用的档位选择器。
   两个 entry 改为顶层 reasoning_effort。kimi-k3 的 NIM 卡片只写了
   "configurable low, high, or max reasoning effort" 而未展示参数,按其第一方
   API 的字段推定,已记入 Unverified。

5. 阿里云 qwen3.8:官方参数表写明「不支持 reasoning_effort 与 thinking_budget
   同时设置,同时设置会报错」,vendor.go 的注释也记了这条,但 applyThinking 在
   ThinkingBudgetTokens > 0 时无条件两个都写。新增 compat 键
   thinking_budget_excludes_effort,冲突时档位优先、预算让路。

3 与 4 在当前默认路径上已经影响到档位;1 和 5 目前够不到(全仓库只有
agent/finalize.go 发 tool_choice="none",且没有任何地方给 ThinkingBudgetTokens
赋值),属于修掉的埋雷。2 是默认路径上的实际错误。

新断言做过反向验证:数据回滚到修改前,7 条 vendor 断言与新增的 golden 用例全部
报错。
2026-09-21 00:05:11 +08:00
..
01-tenant-auth.md fix(security): close the low-severity shared-space findings (#3416) 2026-09-18 22:58:46 +08:00
02-knowledge-base.md fix(security): close cross-workspace access paths in shared spaces (#3408) 2026-09-18 22:00:02 +08:00
03-document-parsing.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
04-chunking.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
05-retrieval-engines.md fix(milvus): support multilingual BM25 retrieval (#3235) 2026-09-14 19:18:24 +08:00
06-models.md fix(models): correct four vendor request shapes that contradict the docs (#3489) 2026-09-21 00:05:11 +08:00
07-agent.md fix(security): close the low-severity shared-space findings (#3416) 2026-09-18 22:58:46 +08:00
08-mcp.md fix(security): close cross-workspace access paths in shared spaces (#3408) 2026-09-18 22:00:02 +08:00
09-knowledge-graph.md refactor(agent): consolidate knowledge retrieval tools into search_knowledge, read_document, list_documents (#3381) 2026-09-18 14:12:22 +08:00
10-datasource.md feat(datasource): add DingTalk Docs integration with safe sync and localization (#2313) 2026-09-14 19:53:29 +08:00
11-web-search.md refactor(agent): consolidate knowledge retrieval tools into search_knowledge, read_document, list_documents (#3381) 2026-09-18 14:12:22 +08:00
12-im-integration.md fix(security): close the low-severity shared-space findings (#3416) 2026-09-18 22:58:46 +08:00
13-embed-channel.md fix(security): close the low-severity shared-space findings (#3416) 2026-09-18 22:58:46 +08:00
14-wiki.md fix(security): close cross-workspace access paths in shared spaces (#3408) 2026-09-18 22:00:02 +08:00
15-evaluation.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
16-observability.md feat(audit): 知识库活动记录对 API Key 操作额外显示 Key 名称 (#3405) 2026-09-18 19:39:34 +08:00
17-faq.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
18-chat-experience.md feat(artifacts): let users delete generated files (#3447) 2026-09-20 13:01:46 +08:00
19-storage-backends.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
20-platform-admin.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
21-file-access.md fix(security): close the low-severity shared-space findings (#3416) 2026-09-18 22:58:46 +08:00
22-skills-sandbox.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00
23-memory.md docs: update website guides and improve writing style (#3095) 2026-09-08 10:54:41 +08:00